We compared DNA sequence promoter patterns of protein-coding and lncRNA genes. We found that A/T-rich mono-, di- and tri-nucleotide patterns are enriched at the promoters of lncRNA genes, relative to the promoters of protein-coding genes (“differentially enriched at lncRNA promoters”) (Table S1). CpG-derived mono-, di- and tri-nucleotide patterns are overrepresented in promoters of protein-coding genes. This result is broadly consistent with the observation that AT-rich promoters demonstrate lower expression but higher tissue specificity [26], properties known to define lncRNA promoters [4]. CG-skew, a feature of protein-coding gene promoters, is significantly reduced in lncRNA gene promoters, while AT-skew is almost depleted (Figure 1a–b). Figure 1c shows that word commonality score (Text S1 Methods section) is decreased around the transcriptional start sites (TSS) of lncRNA genes, although this depletion is stronger around TSSs of protein-coding genes, suggesting that lncRNA gene regulation, in contrast to protein-coding genes, is less driven by unique recognition sequences.
